shmem4py: OpenSHMEM for Python

Marcin Rogowski; Lisandro Dalcin; Jeff R. Hammond; David E. Keyes

doi:10.21105/joss.05444

shmem4py: OpenSHMEM for Python

Marcin Rogowski, Lisandro Dalcin, Jeff R. Hammond, David E. Keyes

Research output: Contribution to journal › Article › peer-review

Abstract

shmem4py brings the Partitioned Global Address Space (PGAS) programming model to Python by exposing the functionality of the OpenSHMEM Application Programming Interface (API) specification. The feature set includes one-sided communication, shared memory access, atomic memory operations, and collective operations. The Python implementation of shmem4py emphasizes using NumPy arrays, providing convenient access to the symmetric memory allocations central to OpenSHMEM’s programming model. Thanks to Python’s versatility and OpenSHMEM implementations’ focus on performance, shmem4py offers a seamless experience on a variety of hardware, from laptops to supercomputers, and for a wide range of applications and users. shmem4py API grounds in OpenSHMEM 1.5 specification; however, it also supports legacy 1.4 implementations.

Original language	English (US)
Pages (from-to)	5444
Journal	Journal of Open Source Software
Volume	8
Issue number	87
DOIs	https://doi.org/10.21105/joss.05444
State	Published - Jul 19 2023

Bibliographical note

KAUST Repository Item: Exported on 2023-07-24
Acknowledgements: The authors thank the KAUST Supercomputing Laboratory for their computing resources.

Access to Document

10.21105/joss.05444

https://repository.kaust.edu.sa/bitstream/10754/693176/1/10.21105.joss.05444.pdf

Cite this

@article{a473131d91824d189d0b0ade16ced09f,

title = "shmem4py: OpenSHMEM for Python",

abstract = "shmem4py brings the Partitioned Global Address Space (PGAS) programming model to Python by exposing the functionality of the OpenSHMEM Application Programming Interface (API) specification. The feature set includes one-sided communication, shared memory access, atomic memory operations, and collective operations. The Python implementation of shmem4py emphasizes using NumPy arrays, providing convenient access to the symmetric memory allocations central to OpenSHMEM{\textquoteright}s programming model. Thanks to Python{\textquoteright}s versatility and OpenSHMEM implementations{\textquoteright} focus on performance, shmem4py offers a seamless experience on a variety of hardware, from laptops to supercomputers, and for a wide range of applications and users. shmem4py API grounds in OpenSHMEM 1.5 specification; however, it also supports legacy 1.4 implementations.",

author = "Marcin Rogowski and Lisandro Dalcin and Hammond, {Jeff R.} and Keyes, {David E.}",

note = "KAUST Repository Item: Exported on 2023-07-24 Acknowledgements: The authors thank the KAUST Supercomputing Laboratory for their computing resources.",

year = "2023",

month = jul,

day = "19",

doi = "10.21105/joss.05444",

language = "English (US)",

volume = "8",

pages = "5444",

journal = "Journal of Open Source Software",

issn = "2475-9066",

publisher = "Open Journals",

number = "87",

}

TY - JOUR

T1 - shmem4py: OpenSHMEM for Python

AU - Rogowski, Marcin

AU - Dalcin, Lisandro

AU - Hammond, Jeff R.

AU - Keyes, David E.

N1 - KAUST Repository Item: Exported on 2023-07-24 Acknowledgements: The authors thank the KAUST Supercomputing Laboratory for their computing resources.

PY - 2023/7/19

Y1 - 2023/7/19

N2 - shmem4py brings the Partitioned Global Address Space (PGAS) programming model to Python by exposing the functionality of the OpenSHMEM Application Programming Interface (API) specification. The feature set includes one-sided communication, shared memory access, atomic memory operations, and collective operations. The Python implementation of shmem4py emphasizes using NumPy arrays, providing convenient access to the symmetric memory allocations central to OpenSHMEM’s programming model. Thanks to Python’s versatility and OpenSHMEM implementations’ focus on performance, shmem4py offers a seamless experience on a variety of hardware, from laptops to supercomputers, and for a wide range of applications and users. shmem4py API grounds in OpenSHMEM 1.5 specification; however, it also supports legacy 1.4 implementations.

AB - shmem4py brings the Partitioned Global Address Space (PGAS) programming model to Python by exposing the functionality of the OpenSHMEM Application Programming Interface (API) specification. The feature set includes one-sided communication, shared memory access, atomic memory operations, and collective operations. The Python implementation of shmem4py emphasizes using NumPy arrays, providing convenient access to the symmetric memory allocations central to OpenSHMEM’s programming model. Thanks to Python’s versatility and OpenSHMEM implementations’ focus on performance, shmem4py offers a seamless experience on a variety of hardware, from laptops to supercomputers, and for a wide range of applications and users. shmem4py API grounds in OpenSHMEM 1.5 specification; however, it also supports legacy 1.4 implementations.

UR - http://hdl.handle.net/10754/693176

UR - https://joss.theoj.org/papers/10.21105/joss.05444

U2 - 10.21105/joss.05444

DO - 10.21105/joss.05444

M3 - Article

SN - 2475-9066

VL - 8

SP - 5444

JO - Journal of Open Source Software

JF - Journal of Open Source Software

IS - 87

ER -

shmem4py: OpenSHMEM for Python

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this